Published as de Laat KF, Reid AT, Grim DC, Evans AC, Kötter R�, van Norden AG, de Leeuw F-E.
Cortical thickness is associated with gait disturbances in small vessel disease. Neuroimage 2011; doi:10.1016/j.neuroimage.2011.08.005.
Abstract
Background and objectives
Although gait disturbances are present in a substantial portion of patients with cerebral small vessel disease (SVD), their pathogenesis has not been clarifi ed as they are not entirely explained by the white matter lesions (WMLs) and lacunar infarcts. The role of cortical thickness in these patients remains largely unknown. We aimed to assess the regions of cortical thickness associated with distinct gait parameters in patients with SVD, and whether these associations were dependent on WMLs and lacunar infarcts.
Methods
MRI data were obtained from 415 subjects with SVD, aged between 50 and 85 years. We assessed cortical thickness using surface-based cortical thickness analysis, and gait using the GAITRite system.
Results
Cortical thickness of predominantly the orbitofrontal and ventrolateral prefrontal cortex, the inferior parietal lobe, cingulate areas and visual association cortices was positively related to stride length. Thickness of the primary and supplementary motor cortices and the cingulate cortex was positively related to cadence, while thickness of the orbitofrontal and ventrolateral prefrontal cortex, anterior cingulate cortex and especially the inferior parietal lobe and superior temporal gyrus was negatively related to stride width. The associations with stride length and width were partially explained by the subcortical WMLs and lacunar infarcts.
Conclusions
Cortical thickness may be important in gait disturbances in individuals with SVD, with different cortical patterns for specifi c gait parameters. We suggest that cortical atrophy is part of the disease processes in patients with SVD.
79
Introduction
Cerebral small vessel disease (SVD), including white matter lesions (WMLs) and lacunar infarcts, is very common in elderly persons.1 SVD is not only associated with cognitive
impairment but also with gait disturbances,116 an often neglected, but nevertheless
important consequence of SVD. However, the anatomical and pathophysiological mechanisms underlying these gait disturbances in SVD remain poorly understood. WMLs and lacunar infarcts have been found to be important factors.116 A number of studies have
furthermore demonstrated that measures of global cerebral atrophy are strongly and independently related to these gait disturbances.37-39,128
Less is known about the specifi c role of cortical atrophy in gait disturbances in individuals with SVD, its spatial distribution and association with subcortical WMLs and lacunar infarcts. Only two studies have investigated the relationship between cortical atrophy and gait performance.38,39 Using volumetric morphometry, they found that cortical volume in
some frontoparietal regions was positively related to gait performance. An alternative MR-based morphometric approach is that of cortical thickness analysis.129 Cortical
thickness methods have been shown to be more sensitive in detecting alterations in cortical morphology than the former volumetric approach,130,131 and have been proven to
be reliable, in terms of spatial localization and magnitude of absolute cortical thickness measurement.132 Moreover, cortical volumetric approaches typically require an a priori
defi nition of a small number of regions-of-interest (ROIs), limiting the possibility of exploratory analyses of cortical regions, whereas surface representations allow a landmark-based assignment of cortical regions to the entire cortex. Because the networks involved in gait are scattered throughout the entire brain in healthy individuals,13 a more
comprehensive and data-driven analysis, such as is provided by this latter approach, is desirable.
It is furthermore not clear whether and to what extent this relationship is (in)dependent of the coexisting WMLs and lacunar infarcts. It could be hypothesized that cortical atrophy in subjects with SVD, and subsequently gait disturbances, may partly be due to WMLs and lacunar infarcts. These lesions may cause disruption of axons traversing the white matter with secondary neuronal degeneration of networks connecting relevant cortical regions, involved in the control of gait. Rosano et al. found that WMLs and lacunar infarcts did not substantially change the regression coeffi cients of these associations, but they assessed WMLs only semi-quantitatively.38,39
Using surface-based cortical thickness analysis, we aimed to identify the association between regional cortical thickness and several spatiotemporal gait parameters in a large group of older adults with cerebral SVD. A secondary objective of this study was to
Cortical thickness and gait disturbances
80
examine whether these associations were dependent on SVD, including WMLs and lacunar infarcts.
Methods
Study population
This study is embedded in the RUN DMC study, a prospective cohort study that investigates risk factors and clinical consequences of brain changes as assessed by MRI among older adults with cerebral SVD, with the development of dementia or parkinsonism as the primary study outcome.117 Participants were recruited from the Department of
Neurology. Recruitment methods and other details of the RUN DMC design are described elsewhere.117 The RUN DMC cohort consists of 503 people, aged between 50 and 85
years, with WMLs and/or lacunar infarct(s) on neuroimaging and acute (transient ischemic attack or lacunar syndrome) or subacute (cognitive, motor or depressive) symptoms of SVD. As suggested for clinical studies, patients were primarily selected on brain imaging features, since clinical symptoms of SVD are more heterogeneous and typically mild at the onset of cerebral SVD.12 Exclusion criteria were: (a) dementia;60 (b) parkinson(ism);61,101
(c) intracranial hemorrhage; (d) life expectancy of <6 months; (e) brain tumor; (f) (psychiatric) disease interfering with cognitive testing or follow-up; (g) recent or current use of acetylcholine-esterase inhibitors, neuroleptic agents, L-dopa or dopa-a(nta) gonists; (h) non-SVD-related WMLs (e.g. multiple sclerosis) (i) prominent visual or hearing impairment; (j) language barrier; (k) MRI contraindications or known claustrophobia. Additional exclusion criteria for this study were: (l) inability to walk 6 m unaided across the GAITRite carpet as this measured 5.6 m (n=4); (m) conditions not related to SVD that affected gait (e.g. joint fusion, severe arthritis or psychogenic gait disorder) (n=11) and (n) territorial infarcts, because they were considered potential confounders and because of methodological issues during the cortical thickness analysis (n=55). Assessment of cortical thickness was furthermore not possible in 18 scans, because of failures in the cortical surface generation process, yielding a sample size of 415 for this study.
All participants signed an informed consent form. The Medical Review Ethics Committee region Arnhem-Nijmegen approved the study.
MRI techniques
All MRI scans of all subjects were acquired on a single 1.5 Tesla scanner (Magnetom Sonata, Siemens Medical Solutions, Erlangen, Germany). The protocol included a 3D T1 magnetization-prepared rapid gradient-echo sequence [repetition time (TR)/echo time
81
(TE)/ inversion time (TI) 2250/3.68/850 ms; fl ip angle 15°; voxel size 1.0x1.0x1.0 mm] and a fl uid-attenuated inversion recovery sequence (TR/TE/TI 9000/84/2200 ms; voxel size 1.0x1.2x5.0 mm, plus an interslice gap of 1 mm).
MRI analysis
WMLs were manually segmented, and the number of lacunar infarcts rated according to a standardized protocol.117 Subsequently, WML volume was calculated.
Cortical thickness was estimated using the Civet pipeline, because this method has proven to be more valid than other used techniques like Freesurfer.133 Firstly, structural T1 images
were aligned to the Montreal Neurological Institute 152 template using a multi-scale, 9-parameter transformation,134 and corrected for nonuniformities using the nonparametric
nonuniformity intensity normalization (N3) method.135 Secondly, the image was segmented
into discrete tissue compartments representing grey matter, white matter, and cerebrospinal fl uid, with partial volume estimates being additionally applied to defi ne deep sulci.129 Thirdly,
deformable surfaces meshes, isomorphic to a sphere, were optimized to fi t the boundaries between white/grey matter and grey matter/cerebrospinal fl uid, respectively, using a multi-scale method called Constrained Laplacian Automated Segmentation with Proximities.129,136 Finally, cortical thickness was calculated as the distance between two
corresponding vertices (i.e. individual spatial points) on the two boundary surfaces. To facilitate intersubject comparisons, individual surfaces were registered to a population template surface obtained through an iterative group template registration algorithm.137
Each cortical surface was parcellated into Brodmann areas (BAs), by a deformation from its representation on the Population-Average, Landmark- and Surface-based atlas
Cortical thickness and gait disturbances
2
82
(PALS-B12)138 (Figure 1) to the template Civet surface. Since all individual surfaces were
previously coregistered to this surface, only one such deformation was necessary. In this parcellation, BAs were derived from manual delineations according to the original cyto- architectonic map of Brodmann.139 Using this scheme, the vascular territories were not
taken into account. Measurement of gait
As reported previously,116 we measured gait velocity (m/s), stride length (m), cadence
(steps/minute) and stride width (cm) using the 5.6m GAITRite system (CIR Systems Inc., Havertown, PA). For this study, we choose velocity, stride length and width, because our previous study demonstrated a clear relationship between these gait parameters on the one hand and WMLs or lacunar infarcts on the other116 and the known association of
velocity with risk of falls.140 As velocity is determined by stride length and cadence, this
latter gait parameter was also included. Other measurements
For assessment of vascular risk factors, structured questionnaires were used and an experienced research nurse measured the blood pressure three times in the supine position after fi ve minutes of rest. The risk factors used were presence of hypertension (mean blood pressure ≥140/90 mmHg and/or use of antihypertensive medications),141
diabetes (treatment with antidiabetic medications), hypercholesterolemia (treatment with lipid-lowering medications) and smoking status. The body mass index was also recorded. We used the Mini Mental State Examination score (range 0 – 30) to assess global cognitive status.64 Functional independence was assessed using the Barthel Index (range 0 – 20).84
Statistical analysis
The baseline characteristics are presented as mean ± standard deviation (SD) and for the positively skewed WML volume, the median and interquartile range was calculated (SPSS statistical software, version 16.0). The quantitative GAITRite parameters were averaged over two walks. When one trial was missing (n=3), the remaining measures were used.
Statistical analysis of cortical thickness was performed using the SurfStat Matlab library,142
by obtaining independent general linear models for cortical thickness values averaged within each subject, for each BA ROI. P-values obtained from these models were corrected for family-wise error (FWE) using the Holm-Bonferroni method. There was one comparison for each BA, and the hemispheres were considered separately (i.e., corrected for the number of areas in one hemisphere rather than both). Also the corrections were made separately for each gait measure. The critical P-values we used were 0.05; these are corrected P-values. The associations between cortical thickness (independent variable)
83
and the gait parameters of interest (gait velocity, stride length, cadence, or stride width) (dependent variable) were tested in two models. The fi rst model was adjusted for age, sex and height. The second model was additionally adjusted for total WML volume (log transformed) and number of lacunar infarcts. We present both regression coeffi cients, to assess the effect size per unit, and standardized beta-values, to compare the associations for the different gait parameters to each other. Cortical maps are presented on the population-based template brain used in the Civet pipeline.
Results
Characteristics
Table 1 shows the demographic, clinical, imaging, and gait characteristics of the study population (n=415). Mean age was 65.1 (SD 8.8) years and 192 (46.3%) were women. The mean cortical thickness was 3.3 (SD=0.2, range 1.8-4.9) mm. Primary motor and somatosensory cortices (BA 1-4) were thinnest (mean thickness 2.2-2.6 mm), whereas the temporal pole and medial temporal lobe were thickest (mean thickness 3.8-4.0 mm). Cortical thickness and gait
Table 2 shows the regression coeffi cients and Figure 2 illustrates the standardized beta-values of the signifi cant associations between the cortical thickness per BA ROI and the different gait parameters. All these regions showed a positive association with gait velocity, stride length, and cadence, and a negative association with stride width, indicating a lower velocity and cadence, a shorter stride length and broader stride width by a decrease of cortical thickness in this study population. Overall, relationships were comparable between both hemispheres. For velocity, signifi cant regions were found bilaterally in BA ROIs involving almost the entire cortex. As expected (since gait velocity is the product of stride length and cadence), these regions correspond to those found for stride length plus cadence.
We found associations between stride length and the cortical thickness of virtually the entire frontal lobe, except for the bilateral primary motor cortices (precentral gyrus, BA 4). The most prominent associations were found in the orbitofrontal (BA 11) and ventrolateral prefrontal cortex (BA 44, 45, 47). For the parietal and temporal lobes these areas were located in the inferior parietal lobe (BA 39, 40), superior temporal gyrus (BA 22, 41, 42), and on the left side the fusiform gyrus (BA 37). A thinner cortical thickness of the visual areas (BA 17-19), as well as the anterior and posterior cingulate areas of the limbic region (BA 23-25, 31-33), was also associated with a shorter stride length.
Cortical thickness and gait disturbances
84
The most prominent associations between a thinner cortical thickness and a lower cadence were located in the left cingulate areas (BA 23-25, 31-33), the bilaterally visual areas (BA 18), and left fusiform gyrus (BA 37). Other signifi cant associations were found in the primary (BA 4) and premotor cortices (BA 6, 8).
Regression analysis between cortical thickness and stride width revealed a signifi cant negative association in both hemispheres in regions largely corresponding to the
Chapter 2.4
Table 1 Characteristics of the study population
Characteristics n=415
Demographic and clinical characteristics
Age, yrs 65.1 (8.8)
Female, no. 192 (46.3)
Height, m 1.7 (0.1)
Body Mass Index, kg/m2 27.1 (4.1)
Subjects with hypertension, no. 296 (71.3) Subjects with diabetes mellitus, no. 59 (14.2) Subjects with hypercholesterolemia, no. 184 (44.3) Smokers, current, no. 61 (14.7) Smokers, former, no. 226 (54.5) Mini Mental State Examination 28.2 (1.6) Barthel Index 19.7 (0.7) Neuroimaging characteristics
Cortical thickness, mm 3.3 (0.2) Grey matter volume, ml 632.9 (69.3) White matter volume, ml 467.6 (66.7) WML volume, ml† 6.4 (3.3; 17.7) Subjects with lacunar infarct(s), no. 122 (29.4) Gait characteristics
Gait velocity, m/s 1.3 (0.3) Stride length, m 1.4 (0.2) Cadence, steps/min 112.1 (10.9) Stride width, cm 10.8 (3.1)
85
Cortical thickness and gait disturbances
2
Table 2 Regions of cortical thickness associated with gait Cortical thickness (mm) Stride length (m) Cadence (steps/min) Stride width (cm) Region (Brodmann area) Regression
coefficients
Regression coefficients
Regression coefficients Left Right Left Right Left Right Frontal lobe
Precentral gyrus (4) - - 7.33 8.23 - - Superior frontal gyrus (6) 0.16 0.16 8.71 8.67 - - Superior frontal gyrus (8) 0.14 0.12 8.91 7.57 - - Superior/middle frontal gyrus (9) 0.17 0.15 7.89 - - - Superior/middle frontal gyrus (10) 0.17 0.17 - - - -2.22 Middle frontal gyrus (46) 0.16 0.21 - - - - Inferior frontal gyrus (44) 0.19 0.26 - - -2.47 -3.50 Inferior frontal gyrus (45) 0.17 0.20 - - -2.80 -2.72 Inferior frontal gyrus (47) 0.11 0.17 - - - -.287 Orbital gyrus (11) 0.22 0.22 - - -2.99 -3.53 Parietal lobe Postcentral gyrus (2) 0.15 0.15 - - - - Postcentral gyrus (3) 0.13 - 8.76 - - - Post-/precentral gyrus (43) 0.15 0.16 - - -2.39 -2.57 Supramarginal gyrus (40) 0.20 0.18 - - -2.79 -3.12 Angular gyrus (39) 0.22 0.13 - - - - Occipital lobe (Peri-)calcarine sulcus (17) 0.15 0.18 - - - -2.66 Lingual gyrus / cuneus (18) 0.15 0.19 9.28 8.97 - - Lateral occipital gyrus (19) 0.19 0.21 - - - - Temporal lobe
Transverse temporal gyrus (41) 0.22 0.24 8.45 8.17 -3.03 -3.66 Transverse temporal gyrus (42) 0.25 0.20 - - -3.83 -3.46 Superior temporal gyrus (22) 0.20 0.15 - - -3.15 -2.97 Middle temporal gyrus (21) 0.11 - - - - - Inferior temporal gyrus (20) - - - - -2.47 -2.69 Fusiform gyrus (37) 0.21 - 9.27 - -2.84 - Limbic system
Subgenual cortex (25) 0.20 0.20 - - - -2.71 Anterior cingulate gyrus (24) 0.18 0.23 10.15 - - -
86
orbitofrontal (BA 11) and ventrolateral prefrontal cortex (BA 44, 45, 47), the cortical areas adjacent to the posterior insula (inferior parietal lobe, superior temporal gyrus, BA 22, 39-42), the inferior temporal gyrus (BA 20) and the left fusiform gyrus (BA 37). Of the limbic lobe, only the dorsal anterior cingulate cortex (BA 32) showed a signifi cant association. As shown in Figure 2, the strength of the associations with stride length (and subsequently velocity) and stride width weakened or even disappeared after adjustment for the subcortical lesions (WMLs and lacunar infarcts), but in most cases a signifi cant association remained. In contrast, the associations between cortical thickness in the frontal lobe and cadence did not markedly change.
Chapter 2.4 Table 2 Continued Cortical thickness (mm) Stride length (m) Cadence (steps/min) Stride width (cm) Region (Brodmann area) Regression
coefficients
Regression coefficients
Regression coefficients Left Right Left Right Left Right Limbic system
Anterior cingulate cortex (32) 0.23 0.21 8.30 9.51 -2.99 -2.78 Anterior cingulate cortex (33) - 0.13 - - - - Posterior cingulate gyrus (23) 0.22 0.18 9.54 - - -2.17 Posterior cingulate cortex (31) 0.20 0.21 11.42 - - - Cingulate cortex (26) - 0.11 - - - - Cingulate cortex (29) 0.11 0.14 - - - - Cingulate cortex (30) 0.15 0.19 - - - -2.82 Parahippocampal gyrus (35) 0.13 - - - - - Parahippocampal gyrus (36) 0.10 0.10 6.41 - - -
Regression coeffi cients represent change in units of the dependent variable by 1 mm decrease of cortical thickness. Only cortical regions with statistically signifi cant associations are presented (P<0.05, FWE- corrected). Adjusted for age, sex and height.
87
Cortical thickness and gait disturbances
2
Figure 2 Regions of cortical thickness associated with gait
Brain regions demonstrating signifi cant (P<0.05, FWE-corrected) associations between cortical thinning and lower gait velocity, stride length, cadence and broader stride width. Effect sizes are illustrated with color-labelled standardized beta-values with scaling from 0.10 to 0.30. Panel A shows models adjusted for age, sex and height. Panel B shows models additionally adjusted for total WML volume and number of lacunar infarcts. Maps are presented on a population template brain. The medial wall and corpus callosum are masked.
88
Discussion
The main fi nding of this study among patients with cerebral SVD was that individuals with a thin cortex performed worse on gait tasks than those with a thicker cortex. There were specifi c anatomical patterns for the various gait parameters. These relationships, with the exception of those observed in the frontal areas with respect to cadence, were partially attributable to subcortical WMLs and lacunar infarcts.
The present study has a number of advantages, including its large sample size, high response rate, and the collection of data from a single scanner and research center. In addition, we used quantitative assessment of cortical atrophy, WMLs, and several gait parameters. Moreover, surface-based cortical thickness analysis provides a more sensitive measure of cortical atrophy than previous approaches130,131 and the use of
surface-based cortical parcellation into BAs permits a more detailed anatomical reference set.138 These advantages allowed us to pursue an approach which was largely data-driven,
employing regional analyses of cortical thickness and gait, rather than hypothesis-driven, necessitating the identifi cation of a small number of pre-defi ned ROIs. As a result, we were able to investigate the relationship of cortical thickness with gait disturbances across the entire cerebral cortex. Finally, we were able to investigate the attribution of SVD in these relations by controlling for WMLs and lacunar infarcts.
At present, our study is cross-sectional, which limits conclusions regarding causality. We cannot prove that a lower cortical thickness precedes gait impairment due to this design. However, future longitudinal studies, such as ours, are needed to examine changes in cortical thickness over time and how they relate to gait. Other markers for cortical neuronal loss could also be used for this purpose, such as the amount of benzodiazepine receptors using positron-emission-tomography.143 Furthermore, the parcellation of the cortical
surface into BAs areas has a number of limitations which are important to consider when interpreting the results described here. Firstly, the use of anatomical landmarks – particularly major sulci – to delineate BAs has been challenged by histological evidence showing that some BAs vary substantially between individuals, both in surface area and location relative to major sulci.144 While this variance is problematic, and can possibly be
addressed through the use of probabilistic atlases, functional localization, or alternate parcellation schemes, the BA parcellation does appear to be highly accurate for some areas, and overlaps substantially with probabilistic representations in more problematic areas.138 Secondly, being derived from cytoarchitectonic criteria, BAs do not necessarily
have a strong correspondence with vasculature. Because SVD, in contrast to large vessel disease, is generally not restricted to these vascular territories, we feel that this has not greatly infl uenced our results. Moreover, we preferred the Brodmann parcellation scheme
89
as it is widely used. In addition, the schemes considering the vascular supply defi ne large territories, which we feel have insuffi cient spatial resolution for our present analyses. Thirdly, given the high variance and apparently unreliable delineation of the insula, this potentially important cortical region was also excluded from our analyses. We furthermore only investigated the cortical thickness of supratentorial structures, although gait is a result of complex networks of both supra- and infratentorial structures.
We found the different gait parameters related to the cortical thickness each at different regions (Figure 2). As gait velocity is calculated from stride length and cadence, we do not separately discuss the results for velocity. Some of the cortical regions – located in the frontal and parietal lobe – of which the thickness related to a shorter stride length and broader stride width, are consistent with that found in another study.39 However, an
interesting difference with this study was observed with regard to the frontal lobe. We could not replicate their fi nding of involvement of the primary motor cortex in a shorter stride length. Instead, we found the strongest associations with stride length and width in